diff --git a/Makefile b/Makefile
index 730358b..20aba3d 100644
--- a/Makefile
+++ b/Makefile
@@ -15,6 +15,7 @@ all: true false cat env hexdump echo
 
 prtest: prtest.o lib/printregs.o lib/hex.o
 sorttest: lib/sort.o
+flagtest: lib/flags.o
 
 .PHONY: test
 test:
diff --git a/flagtest.s b/flagtest.s
new file mode 100644
index 0000000..f2a147a
--- /dev/null
+++ b/flagtest.s
@@ -0,0 +1,35 @@
+
+.global _start
+.extern init_flags
+.extern parse_flags
+
+.data
+
+flags:
+.dword 0 # value / default
+.byte 1 # type: <end>, bool, string, int
+.byte 5 # flag length
+#.short 16 # help length
+.ascii "-help"
+#.ascii "prints help text"
+
+# last flag
+.dword 0
+.dword 0
+
+.text
+
+_start:
+    .option push
+    .option norelax
+    la gp, __global_pointer$
+    .option pop
+
+    mv a0, sp
+    call init_flags
+    la a0, flags
+    call parse_flags
+
+    li a7, 93 # sys_exit
+    li a0, 0
+    ecall
diff --git a/lib/flags.s b/lib/flags.s
index 4e1c66c..aab0c17 100644
--- a/lib/flags.s
+++ b/lib/flags.s
@@ -1,4 +1,12 @@
 
+.option norelax
+.global init_flags
+.global parse_flags
+
+.macro subi dst, src, imm
+    addi \dst, \src, -\imm
+.endm
+
 .data
 
 flags:
@@ -19,6 +27,15 @@ entry_sp: .dword 0
 saved_argc: .dword 0
 saved_argv: .dword 0
 
+.section .rodata
+
+flag_failure_msg: .ascii "flag parsing failed\n\0"
+.equ flag_failure_msg_len, (.) - flag_failure_msg
+unknown_flag_msg: .ascii "unknown flag\n\0"
+.equ unknown_flag_msg_len, (.) - unknown_flag_msg
+
+.text
+
 # when a program starts, it should stash the initial value of the sp register somewhere
 # e.g.
 #
@@ -56,11 +73,13 @@ parse_flags:
     ld  a2, (a2)
 
     tail parse_flags_args
+    # n.b. tail clobbers t1
 
 # a0 - flag definition struct
 # a1 - argc
 # a2 - argv
 parse_flags_args:
+    mv a7, ra
     # TODO: positional arguments
 
     # syntax:
@@ -68,27 +87,53 @@ parse_flags_args:
     #  -flag value      (only non-boolean flags)
     #  -flag=value
 
-    # main loop
+    # if argc <= 1, return
+    li t0, 1
+    bleu a1, t0, .Ldone
+
+    # main loop:
     # for each argument, if it starts with a '-' then see if it matches a flag
-    mv s2, a2
+
+    # skip arg 0
+    addi s2, a2, 8
+    subi a1, a1, 1
 .L_arg_loop:
     ld s1, (s2) # load string pointer
     beqz s1, .L_next_arg  # skip null pointer
     # check if first char is a '-'
-    ld t0, (s1)
-    beqz t0, .Lno # empty string
-    addi t0, t0, -'-'
-    bnez t0, .Lno # not a dash
+    lbu t0, 0(s1)
+    li t1, '-'
+    beq t0, t1, find_flag
+
+.L_next_arg:
+    addi s2, s2, 8
+    subi a1, a1, 1
+    beqz a1, .Ldone
+    j .L_arg_loop
+
+.Ldone:
+    mv ra, a7
+    ret
+
+.Lunknown_flag:
+    mv ra, a7
+    li a7, 64 # sys_write
+    li a0, 2
+    la a1, unknown_flag_msg
+    li a2, unknown_flag_msg_len
+    ecall
+    li a0, -1
+    ret
 
     # TODO: allow --flag
     # TODO: -- should end flag parsing
 
+find_flag:
     mv s0, a0
 .L_flag_loop:
-
     lbu t3, 8(s0) # type
-    lbu t2, 9(s0) # flag length
-    beqz t2, .Lunknown_flag
+    lbu s3, 9(s0) # flag length
+    beqz s3, .Lunknown_flag # len=0 signifies the end of the flag definitions
     call flagcmp
     # return value:
     # 0: no match
@@ -97,42 +142,53 @@ parse_flags_args:
     bnez t0, .Lfound
 
 .L_next_flag:
-    addi s0, s0, 8
-    add s0, s0, t2
+    addi s0, s0, 10
+    add s0, s0, s3
     j .L_flag_loop
 
 
 .Lfound:
-    add t0, s1, t2 # arg+len
-    lbu t1, (t0)
-    beqz t1, 1f # is it a \0 or an =?
-  0:# move s1 past the =
-    addi t0, t0, 1
-    mv s1, t0
-    j 2f
-  1:# is this a boolean flag?
-    li t0, 1
-    beq t0, t3, 2f
+    subi t3, t3, 1
+    add s1, s1, s3 # arg+len
+    lbu t0, (s1)
+    beqz t0, 1f # is it a \0 or an =?
+    # =
+    # move s1 past the =
+    addi s1, s1, 1
+    j handle_flag
+
+  1:# \0
+    # is this a boolean flag? then there's no value
+    beqz t3, handle_flag
     # if not, set s1 to the next arg and advance s2
     ld s1, 8(s2)
     addi s2, s2, 8
+    subi a1, a1, 1
     # TODO: if it's null, that's an error
-  2:
+    # fallthrough
 
+    # s1 now points to the value of the flag
+
+handle_flag:
     # ok now parse the flag
-.Lparse_flag:
-    lbu t3, 8(s0)  # load type
-    li t0, 1
-    beq t0, t3, boolflag
-    li t0, 2
+    #lbu t3, 8(s0)  # load type
+    beqz t3, boolflag
+    li t0, 2-1
     beq t0, t3, strflag
-    li t0, 3
+    li t0, 3-1
     beq t0, t3, intflag
     j .Lbad_flag_def
 
+strflag:
+    # easy, just store the string pointer
+    # TODO: strlen
+    sd s1, 0(s0)
+    j .L_next_arg
+
 boolflag:
     lbu t0, 0(s1)
     # if arg is empty, set value to true
+    # TODO: should -bool and -bool= be treated differently?
     beqz t0, 0f
     # check for "true" and "false"
     li t1, 'f'
@@ -140,63 +196,56 @@ boolflag:
     li t1, 't'
     beq t0, t1, .Lboolcmptrue
     # anything else is an error
-    j .Lboolbad
+    j .Lbadbool
 
 .Lboolcmpfalse:
     lbu t0, 1(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 'a'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 2(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 'l'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 3(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 's'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 4(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 's'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 5(s1)
-    bnez t0, .Lboolbad
-    j 0b
+    bnez t0, .Lbadbool
 
-    sd x0, 0(s0)
+  0:sd x0, 0(s0)
     j .L_next_arg
 
 .Lboolcmptrue:
     lbu t0, 1(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 'r'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 2(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 'u'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 3(s1)
-    beqz t0, .Lboolbad
+    beqz t0, .Lbadbool
     li t1, 'e'
-    bne t1, t0, .Lboolbad
+    bne t1, t0, .Lbadbool
 
     lbu t0, 4(s1)
-    bnez t0, .Lboolbad
+    bnez t0, .Lbadbool
 
-  1:li t1, 1
-    sd t1, 0(s0)
-    j .L_next_arg
-
-strflag:
-    # easy, just store the string pointer
-    # TODO: strlen
-    sd s1, 0(s0)
+  1:li t0, 1
+    sd t0, 0(s0)
     j .L_next_arg
 
 intflag:
@@ -208,7 +257,7 @@ intflag:
     li t2, 10
     lbu t1, 0(s1)
     beqz t1, .Lbadint
-  0:addi t1, t1, -0x30 # '0'
+  0:subi t1, t1, 0x30 # '0'
     bltz t1, .Lbadint
     bge t1, t2, .Lbadint
     mul t0, t0, t2
@@ -219,25 +268,50 @@ intflag:
   1:sd t0, 0(s0)
     j .L_next_arg
 
-
-.L_next_arg:
-    
-
-    ret
-
+.Lbad_flag_def:
 .Lbadint:
-.Lboolbad:
-.Lunknown_flag:
+.Lbadbool:
+    mv ra, a7
+    li a7, 64
+    li a0, 2
+    la a1, flag_failure_msg
+    li a2, flag_failure_msg_len
+    ecall
     li a0, -1
     ret
 
-# t4 - flag
-# t5 - 
-# t6 length
-# t0 - return
+# in: s1 - string to compare, length unknown (null terminated)
+# in: s0 - flag definition
+# in: s3 - flag length
+# out: t0 - match=1  no match=0
+# clobbers t0, t1, t4, t5, t6, ra
 flagcmp:
-    mv t0, zero
+    add t1, s0, 10 # flag
+    add t0, t1, s3 # end of flag
+    mv t4, s1 # str
+  3:lbu t5, (t1)
+    lbu t6, (t4)
+    bne t5, t6, 0f
+    # we don't _need_ to specifically check if t6==\0
+    # since that should be caught by the comparison above
+    # unless the flag contains an embedded \0 (it shouldn't)
+    # but do the check anyway just to be extra safe
+    beqz t6, 0f
+    addi t1, t1, 1
+    addi t4, t4, 1
+    # once we reach the end of the flag, we just need to check
+    # that the argument string ends in the same place
+    bge t1, t0, .Ltail
+    j 3b # loop
+.Ltail:
+    lbu t1, (t4)
+    li t0, '='
+    beq t0, t1, 2f
+    beqz t1, 1f
+  0:mv t0, zero
+    ret
+  2:li t0, 2
+    ret
+  1:li t0, 1
     ret
 
-handle_flag:
-